Probing the coverage and biomass of seaweed is necessary for achieving the sustainable utilization of nearshore seaweed resources. Remote sensing can realize dynamic monitoring on a large scale and the spectral characteristics of objects are the basis of remote sensing applications. In this paper, we measured the spectral data of six dominant seaweed species in different dry and wet conditions from the intertidal zone of Gouqi Island: Ulva pertusa, Sargassum thunbergii, Chondrus ocellatus, Chondria crassiaulis Harv., Grateloupia filicina C. Ag., and Sargassum fusifarme. The different seaweed spectra were identified and analyzed using a combination of one-way analysis of variance (ANOVA), support vector machines (SVM), and a fusion model comprising extreme gradient boosting (XGBoost) and SVM. In total, 14 common spectral variables were used as input variables, and the input variables were filtered by one-way ANOVA. The samples were divided into a training set (266 samples) and a test set (116 samples) at a ratio of 3:1 for input into the SVM and fusion model. The results showed that when the input variables were the normalized difference vegetation index (NDVI), ratio vegetation index (RVI), Vre, Abe, Rg, Lre, Lg, and Lr and the model parameters were g = 1.30 and c = 2.85, the maximum discrimination rate of the six different wet and dry states of seaweed was 74.99%, and the highest accuracy was 93.94% when distinguishing between the different seaweed phyla (g = 6.85 and c = 2.55). The classification of the fusion model also shows similar results: The overall accuracy is 73.98%, and the mean score of the different seaweed phyla is 97.211%. In this study, the spectral data of intertidal seaweed with different dry and wet states were classified to provide technical support for the monitoring of coastal zones via remote sensing and seaweed resource statistics.
Keywords: dominant species; intertidal zone; machine learning; spectral analysis; spectral classification.